With about 40% of the world’s population voting in elections this year, manipulation of information using AI poses a huge threat to democracy. State and nonstate actors are developing strategies to restrict AI-generated audio, images and videos.
Media Monitoring Africa (MMA) and the Electoral Commission of South Africa (IEC) have signed a framework of co-operation with social media platforms Meta, TikTok and Google that commits them to counter false information in the run-up to the May 29 election. “It is unusual to get them all to agree,” says MMA director William Bird. In practice it means those platforms agree to deal quickly with complaints from MMA or the IEC. “They have all the power,” says Bird.
Wits University journalism professor Franz Krüger says society needs to think hard about the implications of AI. “The social fabric relies on a flow of reliable information at a most basic level. It is polluted already, but if it gets worse we really are in trouble.”
Bird gives an example of false AI information: “What we are likely to get at some point is a voice note of [President Cyril] Ramaphosa talking to chief electoral officer Sy Mamabolo, saying something like, ‘Listen, comrade, we need to make sure that we don’t drop below 50%, what steps are you taking to do this?’ And the voice note will be so accurate that people won’t be sure if it is real or not.”
MMA last month launched its guidelines for political parties and media organisations using generative AI. South Africa was also one of eight countries invited to take part in the Forum on Information & Democracy (FID), initiated by Reporters without Borders at the UN in February. The FID country report adopts a human rights-based approach to policy and legislation and advocates co-regulation and self-regulation rather than direct, strict government regulation.
“Today’s effective regulators will rely on sound policy principles, tried-and-tested institutional wisdom and a vanguard spirit,” says South African team leader Dimitri Martinis.
Our approach is to start to build in democratic and rights-based principles on the use of AI
— William Bird
MMA lawyer Tina Power adds: “AI is part of our reality. Instead of panicking or denying its existence or trying to overregulate it and prevent it from being a useful and enabling tool, we wanted to find a way to infuse it into our different spaces in a way that is transparent and rights-based. This ultimately affects the decisions that we will be taking at the end of May.”
The EU has produced the first law to regulate AI. One of the problems with developing regulations, says Krüger, is that “the field is moving so quickly. You have to do what you can and be ready to change.”
Overregulation or incorrect regulation can cause harm, says Power. “If the rules are too severe or if the definitions are too vague and we don’t fully understand what we are dealing with, the results can be dire, particularly in the context of an election where access to information and the free flow of information are vital.”
MMA advocates transparency in the use of AI, the imperative for human oversight, innovation without harm, privacy and confidentiality, eliminating algorithmic bias, co-operation, media diversity and the need for literacy training in AI.
Working with the IEC, MMA has called on the media and political parties to include these principles in their manifestos or guidelines.
“Our approach is to start to build in democratic and rights-based principles on the use of AI,” says Bird. “We want political parties and media that are transparent, accountable and intentional about how they are using AI.”
At a recent event attended by MMA and the IEC, information regulator chair Pansy Tlakula said 13 of the largest political parties had failed a Promotion of Access to Information Act compliance test. Bird says: “We are talking about transparency and accountability around AI, but political parties can’t even comply with a law that was passed 20 years ago. Political parties should be aware that people are watching them and that we expect them to apply the same standards in electoral conduct through their use and application of AI.”
MMA has been working with the IEC’s national party liaison committee to introduce parties to the guidelines and encourage agreement on a code of conduct.
Izak Minnaar of the South African National Editors’ Forum says: “It is the first election where all kinds of AI tools are readily available and ready to use by bad actors and good actors. As is always the case, bad actors are often ahead of the game.”
Would you like to comment on this article?
Sign up (it's quick and free) or sign in now.
Please read our Comment Policy before commenting.